³Ô¹ÏÍøÕ¾

Powering the AFL talent pool through predictive modelling

La Trobe University

Identifying talent in Australian Rules Football has taken a step forward with new research by La Trobe University highlighting the power of predictive player data on draft outcomes.

Researchers, within La Trobe’s Rural Health School and the Bendigo Pioneers, aimed to better utilise the breadth of data collected by the talent pathway to improve understanding of the traits associated with drafted players.

Using four years of historical data, predictions were undertaken for the 2021 AFL ³Ô¹ÏÍøÕ¾ Draft, with the traditional method of logistic regression showing limitations in the ability to pinpoint draft-worthy talent. Neural networks, however, emerged as a game-changer.

Results indicated that draft classification is possible with varying degrees of accuracy, dependent on the available data and the modelling employed. Compared to logistic regression, which tends to be very good at predicting non-drafted players, neural networks were far superior at accurately predicting drafted players across positional groups and data configurations.

Lead researcher and sports scientist Jacob Jennings said the findings show potential for increased objectivity in recruitment decisions and may assist coaching staff to strategically prescribe training that could influence draft success.

“Using these types of models, recruiters can iteratively analyse player draft status as the season progresses,” Jacob said.

“Players could be identified as being drafted early in the season but not appear later, signifying a drop in performance or that they have been outperformed by others.

“In an ideal scenario, the player would continually feature in iterations, or progressively move into predictions as the season progresses.”

The AFL’s elite-junior talent pathway is in place to develop players for future transition into the league.

By gaining deeper insights into the characteristics that set drafted players apart, the lead researcher said development programs could be tailored to maximise each player’s potential and, ultimately, enrich the league with a new generation of elite talent.

“These findings are encouraging and show that we’ve just touched the surface in terms of what can be achieved with the data that is available in the AFL’s talent pathway,” Jacob said. Link to paper:

/Public Release.